High voltage buck converter inductor running too hot

Thread Starter

mostmatt

Joined Jul 31, 2019
4
I'm designing a buck converter with 1500V input voltage, 320V 5A output and I'm having trouble with my inductor running too hot. Circuit is a completely standard buck converter topology with schottky diode built around the TL494. I'm only testing at 1.25A output for the time being (so it's running in discontinuous mode at least according to simulation). My first attempt used a 470uH iron powder toroidal inductor with a 7A rms current rating (ATCA-08-471M) running at 100kHz, and within 10 seconds of operation the enamel on the inductor wire was smoking. I switched to a 250uH inductor with a 10A current rating and doubled the frequency to 200kHz and it seemed about the same. I also tried doubling the frequency with the original inductor, and with both inductors in series and nothing even gets it into the ballpark in terms of power dissipation. The maximum run-time is around 30 seconds before inductor temp is > 150C.

Due to the high voltage I am somewhat limited with diagnostic ability but I can verify that the waveform to the IGBT gate looks good, the IGBT and diode are running cool (enough). Ripple current in this application is not critical, and cost is essentially no object. I just need to get the inductor power dissipation down to something reasonable. What's going on? Is it core losses? Based on simulation and my experiments, I don't believe I'm saturating the inductor. 200kHz is really pushing it as it is, I would like to be at 100kHz for the final design.
 

Thread Starter

mostmatt

Joined Jul 31, 2019
4
Its certainly possible but I'm not seeing or hearing any arcing. I forgot to mention I'm actually running it on 1000v for the time being. Are there special inductors recommended for high-voltage applications? Most datasheets I've seen say hi-pot tested to 2500V or similar with no recommended maximum operating voltage listed.
 
Hi,

Do you have any calculations? Assuming CCM, for Vin = 1500V, Out = 320V, 5A, 1600W, assuming 85% efficiency we get a duty cycle about 0.21. If your switching frequency is 100 kHz I'm expecting your inductor (470 uH) ripple current to be about 5.3A, that sounds "OK". But.. in DCM the peak current will be higher, so higher power loss.
 

Thread Starter

mostmatt

Joined Jul 31, 2019
4
Hi,

Do you have any calculations? Assuming CCM, for Vin = 1500V, Out = 320V, 5A, 1600W, assuming 85% efficiency we get a duty cycle about 0.21. If your switching frequency is 100 kHz I'm expecting your inductor (470 uH) ripple current to be about 5.3A, that sounds "OK". But.. in DCM the peak current will be higher, so higher power loss.
I've only done the basic buck converter calculation to get me my starting point of 470uH.

If I perform a simulation with the parameters you listed above, it of course matches almost perfectly with a duty cycle of 21% and a ripple current of ~5.3A in CCM. If I then reduce the simulated load to my test wattage of 400W leaving everything else the same, the duty cycle becomes ~14.5% and the ripple current drops to about 3.5A in DCM. So with a lower load in DCM, the ripple current should be less. Of course the efficiency will drop since the output power is so much lower, but the total loss in the inductor should still be less as well, correct?

At this point I'm thinking core losses due to high ripple current must be to blame, though the peak current still falls far below the rms current rating of the part.
 

Thread Starter

mostmatt

Joined Jul 31, 2019
4
have you tired reducing the frequency? I have no idea what your resonant frequency is but I wonder if your exceeding it.
I have a 1uF cap on the output so I believe the resonant frequency should be around 7kHz.

I have tried reducing the frequency all the way to around 10khz and there was still a ton of inductor heating likely due to the inductor entering saturation.

Powdered iron is probably not that good at 100 kHz. I suggest that you look toward ferrites or Micrometals Powder Core Solutions.
I purchased a Micrometal Powder toroidal core on your suggestion and will wind an inductor for my next test. The thing is massive at 4" in diameter and should allow me to substantially reduce the ripple current with higher inductance while keeping the I^2R losses fairly low.
 

Wolframore

Joined Jan 21, 2019
1,609
There’s core and wire impedance the calculations are difficult. My guess is you’re hitting the hysterisis causing loss as heat. Might not be a bad idea to reach out to an inductor mfg. those reps love to pass on information. That’s what I would do.
 

DickCappels

Joined Aug 21, 2008
6,185
Another technique to keep your transformer cool is to move from solid wire to several strands of a smaller diameter wire with about the same total cross-sectional area. This reduces the eddy current across the wire perpendicular to the core.

The whole process involves walking back and forth between two points, making several strands in parallel. Then, give one end to your son or daughter and chuck up the other end in an electric drill and turn it on. Twist the set of wires until there are some turns per cm, then at the end, pull to stretch it lightly (so it won't untwist when relaxed). Use that wire to wind your transformer. At high frequencies this will result in a much cooler transformer.
 
Top